Inorganic arsenic (iAs) is a well-known naturally occurring metalloid with abundant hazards to our environment, especially being a human carcinogen through arsenic-contaminated drinking water. The iAs-related contamination is usually examined by a chemical assay system or fluorescence staining technique to investigate iAs accumulation and its deleterious effects. In this work, we present a dual-modality measurement and quantitative analysis methods for the overall evaluation of various dose-dependent iAs-related cytotoxicological manifestations by the combination of the synchrotron-radiation-based scanning transmission soft X-ray microscopy (SR-STXM) and Fourier transform infrared micro-spectroscopy (SR-FTIR) techniques. The gray level co-occurrence matrix (GLCM) based machine learning was employed on SR-STXM data to quantify the cytomorphological feature changes and the dose-dependent iAs-induced feature classifications with increasing doses. The infrared spectral absorption peaks and changes of dose-dependent iAs-induced cells were obtained by the SR-FTIR technique and classified by the multi-spectral-variate principle component analysis (PCA-LDA) method, showing the separated spatial distribution of dose-dependent groups. In addition, the quantitative comparisons of trivalent and pentavalent iAs under high dose conditions (iAsIII_H & iAsV_H) demonstrated that iAsIII_H and its compounds were more toxic than iAsV_H. This method has a potential in providing the morphological and spectral characteristics evolution of the iAs-related cells or particles, revealing the actual risk of arsenic contamination and metabolism.